Automated Image Analysis of Podocyte Desmin Immunostaining in a Rat Model of Sub-Acute Doxorubicin-Induced Glomerulopathy
Drug-induced kidney injury (DIKI) may occur in the absence of remarkable clinical signs of toxicity or clinically relevant increases in routine kidney safety biomarkers . Both blood urea nitrogen (BUN) and serum creatinine are routine biomarkers which lack sensitivity and specificity for nephrotoxicity [1-3]. Remarkable alterations in glomerular filtration rate (GFR) are the best indicator of reduced kidney function. Selected novel, protein-based rat kidney safety biomarkers in urine have been qualified for use in conjunction with BUN, serum creatinine and histopathology to enable monitoring for compound induced tubular and/or glomerular toxicity . In the rat, drug-induced tubular lesions may be predicted by increases in urinary protein levels of kidney injury molecule 1 (KIM-1), albumin, trefoil factor 3 and clusterin. Impairment of tubular reabsorption and alterations or damage of glomeruli may be identified by changes in urinary levels of total protein, cystatin C and B2-microglobulin. MicroRNAs (miRNAs) function at the post-transcriptional level by modifying translation, or promote the cleavage of their target mRNAs . Because miRNAs play a key role in gene regulation and demonstrate high inter-species conservation, these species have the potential to serve as novel indicators of DIKI that may outperform traditional biomarkers; when elevated, miRNAs have the potential to serve as novel genomic indicators associated with renal injury that may outperform BUN and serum creatinine when monitoring for translatable DIKI liabilities .
In an investigative toxicology study, naïve male Sprague Dawley (SD) rats received once weekly bolus intravenous (iv) injections of doxorubicin (Dox) to induce microscopic primary glomerular lesions and secondary tubular lesions. Changes in selected novel, non-invasive, tubular-specific, urinary protein-based DIKI biomarkers associated with progressive Dox (5 mg/kg/dose)-induced kidney injury in the absence of remarkable clinical signs of toxicity was evaluated as described previously [3,5]. The earlier findings showed that urinary albumin was more sensitive and selective than urinary total protein, lipocalin-2, KIM-1 and/or osteopontin (3, 5) as well as microRNA-34c-3p (5) in the prediction of progressive Dox-induced glomerular toxicity with secondary renal tubular toxicity. Recently, next generation diagnostic molecular pathology tools have evolved with the intent to support the assessment of archived tissue specimens [6-8]. We developed a novel computational method to further characterize Dox-induced glomerular lesions based on retrospective quantification of changes in glomerular/adjacent non-glomerular and podocyte-specific injury biomarkers.
Materials and Methods
All in vivo animal procedures were conducted in an Association for Assessment and Accreditation of Laboratory Animal Care International accredited facility under an Institutional Animal Care and Use Committee approved protocol. Standard procedures and conditions for animal care, caging, access to water and food, environment, and room maintenance were used. All other procedures were performed in accordance to laboratory standard operating procedures and/or established laboratory best practices. The data described have not been previously reported and are limited to those most relevant to the development of a novel computational method for retrospective quantification of Dox (7.5 mg/kg/ dose)-induced changes in select invasive podocyte-specific injury biomarkers.
Rat Doxorubicin Nephrotoxicity Model
Naïve male SD rats (Charles River Laboratories, Hollister, CA, USA), 7 to 8 weeks old and weighing 249 to 271 g, were used in this study. Animals were randomly assigned to either a control (Vehicle, Veh) group or test article (Dox) group as shown in Table 1. Doxorubicin hydrochloride (Sigma-Aldrich, St. Louis, MO, USA) was formulated in 0.9% Sodium Chloride/Saline for injection, USP (Baxter, Deerfield, IL, USA). On study days 1 and 8, rats were administered, via a lateral tail vein, a bolus iv injection of either Dox (7.5 mg/kg/ dose) or Vehicle (0.9% Saline). Whole blood was collected after 7 and 10 days (post dose) and at termination to generate serum for traditional chemistry parameter analysis. Urine was collected after 7 days (post dose) only to monitor for treatment-induced progressive changes in total urine output. The scheduled study termination was day 14. All animals were euthanized by asphyxiation with carbon dioxide followed by exsanguination.
Table 1. On study days 1 and 8, rats were administered via a lateral tail vein a bolus intravenous (iv) injection of either Doxorubicin (Dox, 7.5 mg/kg/dose) or Vehicle (Veh, 0.9% Saline).
Siemens Advia 1800 automated chemistry system and reagents (Siemens Corporation, Washington, DC, USA) were used to measure concentrations of serum creatinine, BUN, cholesterol and triglyceride.
Representative Tissue Collection and Routine Histology
At the time of necropsy, left kidneys were macro-dissected to collect representative cortices which were subsequently flash-frozen and stored at -80°C until time of analysis. Additionally, left kidney, liver and heart tissue samples were collected and immersed in 10% neutral buffered formalin for 48 hr. The formalin-fixed left kidney samples were trimmed longitudinally and further processed, embedded in paraffin, sectioned at 4-μm, mounted on glass slides, de-paraffinized and stained with hematoxylin and eosin for subsequent microscopic evaluation. Severity scores were assigned based on the classification of microscopic evidence of tissue injury; the distribution and increased numbers of cells affected were assigned a qualitative severity score: 0 = no abnormality noted, 1 = slight (minimal, <25%), 2 = mild (25-50%), 3 = moderate (>50%) and 4 = severe (>75%).
Kidney Desmin Immunohistochemistry Staining
Formalin-fixed left kidney samples were sectioned at 4 mm, mounted on glass slides, de-paraffinized, and hydrated in distilled water followed by phosphate-buffered saline (PBS) solution. Heat-mediated antigen retrieval with citrate buffer (pH ~6) was performed prior to commencing with the IHC staining protocol. Immunohistochemistry (IHC) staining was performed using the DAKO Autostainer (DAKO Cytomation, Carpinteria, CA, USA). Slides were pretreated with endogenous enzyme blocking solution (3% hydrogen peroxide), avidin/biotin blocking solution and protein blocking solution (10% normal goat serum) to avoid non specific reaction with primary antibody. The kidney sections were incubated in the presence of the primary antibody, rabbit anti desmin monoclonal antibody (EMD Millipore, Billerica, MA, USA) at a dilution of 1:500. Goat anti rabbit biotinylated IgG (EMD Millipore) at a dilution of 1:2000 was used as the secondary antibody.Vascular smooth muscle within the kidney served as an internal positive control tissue. Isotype matched IgG (normal rabbit) matching the concentration of primary antibody was used as a negative control. The immunoreactivity was visualized using the Avidin/Biotin Complex reagent (Vector, Burlingame, CA, USA) and diaminobenzidine (DAB, DAKO) followed by counterstaining with Mayer’s hematoxylin.
Automated Image Analysis
We developed a novel computational method to automatically detect glomeruli to quantify the immunopositive podocyte- specific desmin staining (area and intensity) using VISIOmorph® software (Visiopharm, Horsham, Denmark). IHC stained desmin slides were scanned at a magnification of 40X using NanoZoomer 2.0 HT (Hamamatsu Photonics, Middlesex, NJ, USA) whole slide scanner, and virtual slides were imported into the Visiopharm® software platform. Briefly, a linear threshold classifier, which was pre-trained using hematoxylin/DAB and DAB (HDAB-DAB) feature combined with a mean filter and poly local linear filter, detected glomerular tissues in whole-slide images during scanning with a 1024 x 1024–pixel detection window. Next, boundaries of the detected glomeruli were automatically determined using post processing, including edge detection, smoothing, binning, and morphology operation followed by the quantification of immunopositive area (percentage of desmin positive area) and intensity. On average, approximately 200 glomeruli in the whole cortex, including outer and juxtamedullary cortex, were automatically recognized and analyzed. Results were expressed as mean±SD. Differences between groups were calculated by unpaired t-test. The differences between groups were considered to be statistically significant when p<0.05.
Total RNA (including both mRNA and miRNA) was extracted from representative flash frozen macro-dissected left kidney cortices by homogenization with a TissueLyser (Qiagen, Valencia, CA, USA) followed by purification with an AllPrep DNA/ RNA/miRNA Universal Kit (Qiagen) in accordance with the manufacturer’s instructions. Total RNA purity and concentration was measured using a Nanodrop 8000 spectrophotometer (Life Technologies, Grand Island, NY, USA), and the quality of each RNA sample was determined using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA).
SYBR® Green LNA™ Quantification of MicroRNA-34c-3p
MicroRNA-34c-3p levels were measured using SYBR® Green LNA™ technology (Exiqon, Woburn, MA, USA). The synthetic miRNA UniSp6 (Exiqon) was spiked into aliquots of total RNA from each sample to examine technical reproducibility. Following the manufacturer’s suggested protocol, 20 ng of macro-dissected left kidney cortex total RNA was amplified using the Universal cDNA Synthesis Kit II (Exiqon). Sample analysis by RT-qPCR was performed using miR-34c-3p, UniSp6, and U6 probes; ExiLENT SYBR® Green Master Mix (Exiqon); and ROX Reference Dye (Invitrogen, Carlsbad, CA, USA) per manufacturer’s protocol. Samples were run in triplicate using an ABI PRISM 7900 HT Real-Time PCR Instrument (Life Technologies). Ct values generated were normalized to the endogenous control, U6. Significance (p<0.05) was calculated for normalized Ct values using a Student’s t-test. Fold changes for significantly altered expression between Dox group tissues and Veh group tissues were determined by calculating 2(-ΔΔCt).
TaqMan® Gene Expression Assays
Total RNA (1 μg) was reverse transcribed to cDNA using the High Capacity cDNA Reverse Transcription Kit (Life Technologies), per the manufacturer’s suggested protocol. For each probe, samples and a no template control (water) were run alongside no reverse transcriptase controls in triplicate. Desmin (Des) and podocin (Nphs2) mRNA expression analyses were quantified using the corresponding rat specific TaqMan® assays (Life Technologies) as summarized see notes at Table 2 on the ABI PRISM 7900HT Real Time PCR System (Life Technologies). Beta actin was selected as the endogenous control. Results were expressed as mean±SD. Differences between groups were calculated by unpaired t-test, and considered to be statistically significant when p<0.05. Fold changes for significantly altered expression between Dox group tissues and Vehicle group tissues were determined by calculating 2(-ΔΔCt), as described above.
Table 2. MicroRNA-34c-3p assays and rat-specific mRNA gene expression assays. *Purchased from Life Technologies (Carlsbad, CA, USA). **Purchased from Exiqon (Woburn, MA, USA). Abbreviation: hsa, homo sapiens.
Select group data are presented as mean±SD, except where noted. Where applicable, fold changes (from controls) are indicated. As appropriate, comparison between two groups at the same time point was made by unpaired Student’s t test to reveal statistical significance (indicated as ***p<0.001, ** p<0.01 and *p<0.05) using GraphPad Prism software version 6 for Windows (GraphPad Software, Inc., San Diego, CA, USA).
Clinical Observations and Mortality
Numerous clinical signs of toxicity including ventral staining (4/5), oral discharge (3/5), red pigmented urine (1/5), and gross skin (injection site) reactions (1/5) were observed in the Dox group animals. Due to the severity of progressive clinical signs, 1 of 5 Dox-treated animals was euthanized on study day 7, and low blood volumes (3/5) were observed in the surviving Dox-treated animals which were euthanized on study day 10.
Serum Chemistry and Urine Output
When compared to concurrent controls, concentrations of cholesterol were increased (1.6-fold, p=0.008) in the Dox-treated animals on day 7 (Figure 1A); however, there were no Dox-induced changes in BUN on day 7 (Figure 1C).
When compared to the Dox-induced concentration changes observed on day 7, progressive elevations in cholesterol (2.4-fold, p=0.007, Figure 1B) and BUN (2.2-fold, p=0.0004) were measured in the Dox group on day 10 (Figure 1D). On day 7, mean total urine outputs for the Veh (controls) and Dox groups were 20.8±6.8 mL and 4.0±0.9 mL, respectively. When compared to concurrent controls, total urine volumes were decreased (5.2-fold, p=0.0006) in the Dox-treated animals (individual values not shown).
Figure 1. Changes in traditional serum biochemistry parameters. Significant Dox-induced elevated serum cholesterol concentrations were observed on days 7 (A) and 10 (B). There were no remarkable treatment related changes in blood urea nitrogen (BUN) concentrations on day 7 (C). Significant Dox-induced increases in BUN concentration changes on day 10 (D) when compared to concentrations observed on day 7 (C). Values significantly different from Veh (control) or day 7. Dox-treated animals are indicated as ***p=0.0004 (C), **p=0.007 (B) or **p=0.008 (A).
Amongst the 7.5 mg/kg/dose group animals, gross pathological findings included absence of visceral abdominal adipose tissue (2/5) and decreased body fat (1/5). Primary glomerular lesions and secondary tubular lesions were identified following once weekly administrations of Dox (Figure 2, B-D and Table 3). All other kidney histopathology findings were considered unrelated to treatment because they were either present in controls or were consistent with common background findings in healthy male SD rats (data not shown). There were no Dox-related histopathology findings in the liver or heart.
Table 3. Dox-induced microscopic rat kidney histopathology severity scores (1 = minimal lesion, 2 = mild lesion, and 3 = moderate lesion). There were no kidney histopathology findings amongst the control animals that received Veh (0.9% Saline, data not shown).
Figure 2. (A) Representative kidney photomicrographs. No remarkable histopathology findings were identified in kidneys from controls. Dox-induced primary glomerular lesions and secondary tubular lesions were characterized as moderate multi-focal hyaline casts (intratubular, yellow asterisk); mild multi-focal decrease of (B) Bowman’s space and podocyte expansion (glomeruli, green arrow, minimal multi-focal hyaline casts (intratubular); minimal multi-focal hyaline droplets (within podocytes); minimal multifocal tubular vacuolization with fine hyaline casts (cortico-medullary junction) suggesting protein leakage from damaged glomeruli;minimal multi-focal tubular vacuolization with fine hyaline droplets (cortico-medullary junction, blue arrow with yellow highlight, D); minimal multifocal tubular dilatation and degeneration with individual cell necrosis (cortico-medullary junction, yellow arrow, D); and minimal multi-focal tubular epithelial cell regeneration with mitotic figures (cortex, blue arrow, C). 40x magnification.
Kidney Desmin Immunohistochemistry
To analyze quantitative podocyte desmin immunoreactivity, we have developed a novel computational method to automatically detect glomeruli for subsequent quantification of immunopositive desmin IHC staining area and intensity. The automated HDAB-DAB feature, combined with a mean and poly-local linear filter, trained the Visopharm VISIO morph software to use a linear threshold classifier to recognize the glomeruli within the kidney tissues (Figure 3). This feature is an excellent choice for staining variabiity even amongst weakly stained glomeruli. As a result, this method automatically detected boundaries of the glomeruli in a whole kidney section using post processing, and subsequently quantified the glomerular area, immunopositive area (%) and intensity in the sub-image.
This technique revealed sparse desmin staining in podocytes in kidneys from control group animals (Figure 4A and 4B), while the Dox-treated animals displayed a marked increase in podocyte desmin immunoreactivity (Figure 4C and 4D). Significantly increased whole cortex, podocyte desmin immunopositive area (Figure 5A) and intensity (Figure 5B) was evident in rats treated with Dox (compared with controls, p<0.001 and p<0.01, respectively). It has been reported in the rat puromycin aminonucleoside nephrosis model [6, 8] that podocyte injury, evaluated by desmin immunostaining, occurred in juxtamedullary glomeruli before cortical glomeruli. Consequently, we also used automated assessment to analyze the podocyte desmin-positive area and intensity in outer cortical and juxtamedullary glomeruli in Dox treated rats. The results revealed a marked difference in podocyte desmin immunoreactivity between outer and juxtamedullary glomeruli (Figure 6A and 6B).
Figure 3. Workflow of the developed, automatic computational image analysis method. A whole kidney section image was scanned with a detection window, and candidate sub-images were cropped. Using extracted HDAB-DAB features combined with a mean and poly local linear filter, the pre-trained linear threshold classified the candidate cropped sub-images containing glomerular and non-glomerular tissues. As a result, the algorithm automatically recognized glomeruli in a whole kidney section and quantified the glomerular desmin-immunopositive area and intensity thereafter.
Figure 4. Representative photomicrographs of desmin immunohistochemistry staining in podocytes from rats administered Dox showed enhanced podocyte desmin immunoreactivity at the edge of the glomerular tuft (typical anatomic position of podocytes, C and D) as compared to controls (A and B). 10x and 40x magnifications.
Figure 5. Automated, quantitative image analysis of whole corticalglomeruli in desmin immunostained kidneys from Veh-treated and Dox-treated rats. Glomerular desmin positive area (A) and desmin intensity (B) are shown. Statistically significant differences are indicated by ***p<0.001 and **p<0.01.
Figure 6. Desmin immunoreactivity in outer cortical (green arrows) and juxtamedullary (red arrows) glomeruli of Dox-treated rat (A). High podocyte desmin immunoreactivity is shown in juxtamedullary glomeruli. Automated assessment to analysis in the same area (B). Glomeruli are recognized in green color and desminpositive labeled podocytes are in red color. More desmin labeled podocytes are located in juxtamedullary glomeruli. Automated, quantitative image analysis (C-D). Glomerular desmin-positive area (C) and desmin intensity (D) in both outer cortex and juxtamedullary cortex are shown. Statistically significant differences are indicated by **p<0.01 and *p<0.05.
Statistically significant increases in desmin positive area (p<0.01, Figure 6C) and in desmin intensity (p<0.05, Figure 6D) were observed in juxtamedullary glomeruli as compared with outer cortical glomeruli. These findings indicate that podocyte desmin immunoreactivity identified using the novel automated image analysis method could be leveraged as a more sensitive approach for characterizing Dox-induced glomerulopathy. This method may also be adapted to optimally characterize other sub-acute glomerular injury models.
Kidney Genomic Analysis
When compared to controls, significant increases in Des (4.63-fold, p<0.005) and Nphs2 (2.18-fold, p<0.05) gene expression were observed in macro-dissected kidney cortices from the Dox-treated group animals, while miRNA-34c-3p was also significantly increased (3.54-fold, p<0.05) in the Dox-treated animals (Figure 7).
Figure 7. Dox-induced significant upregulation of Des mRNA and Nphs2 mRNA expression and miR-34c-3p enrichment in macrodissected rat kidney cortices. Data shown represents the mean fold change±SD. Statistically significant differences are indicated by *p<0.05 and **p<0.005.
Traditionally, electron microscopy is used to characterize morphologic damage to podocytes . In the present study, progressive Dox-induced increases in cholesterol concentrations on days 7 and 10 occurred in the absence of histologic evidence of liver injury on day 10, while increases in the traditional, non-specific renal safety biomarker, BUN were first observed on day 10 in male SD rats following once-weekly dosing of Dox (7.5 mg/kg). Dose related serum lipidemia was observed in this study, whereas increased serum cholesterol concentrations are a hallmark minimally invasive indicator of progressive Dox nephrotoxicity . This phenomenon serves as a compensatory response to intravenous doses at 4 mg/kg ( and 5 mg/kg  Dox induced proteinuria in rats which results in decreased urine volume (as was also observed in this study as early as approximately day 7 amongst the 7.5 mg/kg Dox-treated animals) and decreased colloid osmotic pressure .
We developed a novel computational method to automatically detect glomeruli and to quantify the glomerular positive (area and intensity) desmin staining as an indicator of nephron specific injury. Significant desmin positive immunolabeling (as a sensitive indicator of podocyte injury) in formalin fixed, paraffin embedded kidney histologic sections and concurrent significant increases in podocyte marker (Des mRNA and Nphs2 mRNA) expression in flash frozen macrodissected kidney cortices were observed in Dox-treated rats (when compared to controls). Additionally, increased enrichment of the novel glomerular injury marker, miRNA-34c-3p  was detected in flash frozen macro-dissected kidney cortices from rats that received Dox. These findings suggest that our novel automated quantitative image analysis of enhanced podocyte desmin IHC staining could serve as an alternative approach to electron microscopy when further characterization of drug-induced alterations in podocyte morphology is warranted.
Taken together, we demonstrated a fully automated image analysis method for analyzing enhanced desmin immunoreactivity in histologic kidney sections in association with Dox induced glomerulopathy in male SD rats. This time saving molecular pathology enabling tool could augment histopathological evaluations in preclinical studies as well as clinical settings. Significant desmin immunolabeling appeared as a sensitive marker for early podocyte alterations and for activated podocytes. These changes appeared concurrent to significant increases in the expression of the protein-encoder gene Des in macrodissected kidney cortices from Dox-treated rats. Alterations in the expression levels of miRNA-34c-3p and Nphs2 in kidney cortices support the consideration of these RNA species as genomic indicators of Dox related primary glomerular damage and secondary tubular injury, respectively. These findings also suggest that further investigations to understand the role of miRNA related gene targets in DIKI is warranted. For example, a prospective time course study with an exemplar glomerular toxicant in rats, particularly designed to include temporal sampling of serum and urine for the evaluation of traditional and novel kidney safety biomarkers. Additionally, laser capture microdissected glomeruli and adjacent non-glomerular cells could reveal the utilities of miRNA-34c-3p expression and biologically relevant downstream glomerular-specific markers in association with glomerular toxicant-induced lesions.
We would like to thank Antonio Guy, Matthew Nemec and Michele Rizzolio for conducting the in life phase of this study. Additional thanks to Benjamin Freiberg, Ph.D., Technical Services, Visopharm, Broomfield, CO, USA for his assistance with selected technical issues regarding the digital pathology analysis method. We would also like to thank the Renal Integrated Safety Assessment team within Janssen Research & Development, L.L.C. for their ongoing support to ascertain novel renal biomarkers within specific contexts of use to enable proprietary drug development programs to progress more rapidly to deliver safer medicines intended to meet unmet medical needs.
1.Dieterle F, Sistare F, Goodsaid F, Papaluca M, Ozer JS et al. Renal biomarker qualification submission: a dialog between the FDA EMEA and Predictive Safety Testing Consortium. Nat Biotechnol. 2010, 28(5): 455-462.
3.McDuffie JE, Sonee M, Ma J, Almy F, Liu X, Xuejun Liu et al. Feasibility of protein biomarkers in the prediction of subclinical doxorubicin nephrotoxicity in male sprague dawley rat. J Mol Biomark Diagn. 2014, 5: 165.
5.Church RJ, McDuffie JE, Sonee M, Otieno M, Liu X et al. MicroRNA 34c 3p is an early predictive biomarker for doxorubicin-induced glomerular injury progression in male sprague dawley rats. Toxicol Res. 2014, 3: 382-394.
6.Herrmann A, Tozzo, E, Funk J. Semi-automated quantitative image analysis of podocyte desmin immunoreactivity as a sensitive marker for acute glomerular damage in the rat puromycin aminonucleoside nephrosis (PAN) model. Exp Toxicol Pathol. 2012, 64(1-2): 45 49.
8.Kakimoto T, Okada K, Hirohashi Y, Relator R, Kawai M et al. Automated image analysis of a glomerular injury marker desmin in spontaneously diabetic Torii rats treated with losartan. J Endocrinol. 2014, 222(1): 43-51
9.Joles JA, Bijleveld C, van Tol A, Geelen MJ, Koomans HA. Plasma triglyceride levels are higher in nephrotic than in analbuminemic rats despite a similar increase in hepatic triglyceride secretion. Kidney Int. 1995, 47(2): 566-572.