Introduction

Podocytes are highly specialised, terminally differentiated epithelial cells located at the urinary aspect of the glomerular basement membrane (GBM) in the kidney. They are critical in maintaining glomerular filtration barrier integrity thus preventing the development of albuminuria, an independent risk factor for end-stage-kidney-disease [ESKD] (1), cardiovascular morbidity (2) and death. Therefore, understanding the critical signalling pathways necessary for maintaining podocyte health is desirable.

The insulin/insulin-like growth factor (IGF) signalling axis regulates many aspects of growth and metabolism: insulin signalling is predominantly associated with metabolism and IGF signalling with growth and cell division (3). However, insulin, IGF1 and their cognate receptors (the insulin receptor [IR] and insulin-like growth factor receptor [IGF1R]) share significant structural homology and several common downstream signalling nodes (4). The IR and IGF1R can bind each other’s ligands and can act redundantly in some contexts (5, 6). The constitutive global (whole body) loss of either the IR or IGF1R in murine knockout models is lethal (7, 8) suggesting a limited ability for these receptors to compensate for each other in every cell type. However, it is now clear that the insulin / IGF axis has different roles in different cell types. Several mouse models with tissue-specific dual knockout of both the IR and IGF1R show differential consequences occurring in a cell type dependent manner (913).

Insulin signalling to the podocyte via the IR is critical for podocyte function (1417) and it is apparent that IGF signalling is also important for podocyte homeostasis (1820). Recently we have shown that loss of podocyte IGF1 signalling can have both detrimental and beneficial effects dependent upon the level of IGF1R suppression: whilst near total loss of podocyte IGF1R expression results in severe mitochondrial dysfunction associated with cell death, partial receptor inhibition promotes podocyte survival in conditions of cellular stress (21).

Here we show that combined simultaneous loss of both the IR and IGF1R in podocytes is physiologically highly detrimental and present proteomic and transcriptomic analyses indicating that the IR/IGF1R hormonal signalling axis is crucial in controlling spliceosome assembly and RNA processing in this cell.

Methods

Mouse models

Mice in which exon four of the IR (7) and exon three of the IGF1R have been flanked by loxP sites (8) were crossed with podCre mice (22). Progeny with the genotype podCre +/- IR wt/fl IGF1Rwt/fl was then crossed with IRfl/fl IGF1Rfl/fl mice to generate podocyte specific IR/IGF1R knockdown animals from embryonic day 12 (podDKO) (Supp Figure 1A). Cre negative littermates served as controls. All mice were on a mixed genetic background including contributions from 129/Sv, FVB and C57BL/6. Both sexes were studied, and no phenotypic differences observed.

podDKO mice develop a renal phenotype by 24 weeks

A. and B. Body weight (A) and blood glucose (B) are not significantly different in podDKO mice at 24 weeks compared with littermate controls.

C. Urinary Albumin Creatinine ratio (uACR) is significantly increased in podDKO mice at 24 weeks. Unpaired t-test, **p<0.01, n=7-8 mice per group.

D. Images and quantification of PAS staining shows tubular protein casts (indicated by arrows) and glomerulosclerosis in podDKO mice. Scale bar=25 μm. Unpaired t-test *p<0.05.

E. Masson’s trichrome staining shows increased fibrosis (blue staining) in podDKO mice at 24 weeks.

Scale bar=25 μm

F. Transmission Electron Micrograph (TEM) images of glomerular filtration barrier (GFB) (scale bar upper panel=5 μm, lower panel=500 nm) show ultrastructural damage to the GFB in podDKO mice with significantly increased foot process width. Unpaired t test **p<0.01

G. Immunofluorescent staining and quantification of WT1 in podDKO and Cre negative control mice at 24 weeks of age shows significant reduction in % podocytes per glomerulus in podDKO mice. Nuclei counterstained with DAPI. Scale bar=100 μm. Unpaired t-test ****p<0.0001, ≥8 glomeruli analysed per mouse, 3 mice per group. Scale bar=100 μm.

Transgenic mouse work was carried out in accordance with the University of Bristol’s institutional guidelines, and procedures approved by the United Kingdom (UK) Home Office in accordance with UK legislation. (Home Office Protocol numbers PPL 3003394 & PPL 2012285).

Urinary albumin and creatinine measurements

Albumin and creatinine levels in spot collections of mouse urine were measured using a mouse-specific albumin ELISA (Universal Biologicals, Cat# E90-134) and creatinine companion kit (Exocell, Cat# 1012), following the manufacturer’s methodology.

Histology and periodic acid-Schiff staining

Kidneys were fixed in 10% buffered neutral formalin, further processed and paraffin embedded. 3 μm sections were cut and stained using a Periodic acid-Schiff staining kit (Sigma) according to the manufacturer’s instructions. Tissues were imaged using a Leica DN2000 microscope and micrographs taken using Leica Application Suite software. Image analysis was performed using ImageJ; all images were contrast enhanced using the same parameters.

Glomerulosclerosis was scored for each glomerulus as follows: 0=normal glomeruli; 1=up to 25% involvement; 2=25-50% involvement; 3=50-75% involvement; 4=over 75% involvement. The glomerulosclerosis index was calculated as described previously (23) and according to the formula [(1xN1)+(2xN2)+(3xN3)+(4xN4)]/(N0+N1+N2+N3+N4) were Nx is the number of glomeruli with each given score.

Electron microscopy

Tissues for electron microscopy were fixed in 0.1 M sodium cacodylate, 2% glutaraldehyde, and imaged on a Technai 12 transmission electron microscope. Average slit diaphragm, foot process and glomerular basement membrane width were calculated using ImageJ assessing at least 20 regions of glomerular basement membrane from at least 2 glomeruli per mouse.

Lentiviral transduction of a conditionally immortalised IR/IGF1R floxed podocyte cell line

Kidneys were isolated from IR fl/fl IGF1Rfl/fl mice and used to make a temperature-sensitive SV40 conditionally immortalised podocyte cell line using techniques as described previously (24) (Figure S2). Podocytes were cultured at 33°C and when 50% confluent were thermo-switched to 37°C and incubated for a further 7 days before transduction with a lentivirus expressing Cre recombinase as described previously. Transduction was in RPMI media with hexadimethrine bromide (Sigma) at 4 μg/ml and the virus used at a multiplicity of infection of 1. Following a 24-hour incubation, the lentivirus was removed and replaced with fresh media. Cells were incubated for a further 3-7 days before imaging and protein extraction.

To determine ciDKO cell number, cells were washed 3 times in PBS, the nuclei stained with Hoechst (Sigma) at 1 μg/ml and imaged using an IN Cell analyser 2200 (GE Healthcare), and data analysed using IN Cell analyser Developer software (GE Healthcare).

Cell culture

Conditionally immortalised podocyte cell lines were cultured as described previously (25). For acute insulin and IGF1 stimulation, cells were serum starved for 4 hours, then 10 nM and 100 nM of insulin or 10 ng/ml and 100 ng/ml IGF1 was applied to the cells for 10 minutes. Spliceosome inhibitor treatment was carried out using pladienolide B at 0-100 nM for 48 hours before determination of cell number.

Western blotting

Cultured cells were lysed in radioimmunoprecipitation assay buffer supplemented with protease and phosphatase inhibitors (Sigma). 10-30 μg of protein was resolved by electrophoresis and then transferred to a polyvinylidene difluoride membrane (Millipore). Membranes were blocked in TRIS-buffered saline with 0.1% tween 20 and 5% BSA for 1 hour and then incubated overnight with primary antibody at a dilution of 1:1000. Membranes were washed before incubation with horseradish peroxidase conjugated secondary antibody (Sigma). Immunoreactive bands were visualised using Clarity ECL Western blotting substrate (Biorad) on a GE AI600 imager. Densitometry was performed using ImageJ software.

Proteomic analysis

ciDKO podocytes (3 days after Cre lentiviral transduction and before significant cell loss) were lysed in RIPA buffer and subjected to LC-MS/MS using isobaric TMT labelling as described previously (26, 27). n=9 ciDKO and control podocytes were analysed. Wild-type mouse podocytes served as controls.

The data output from Proteome Discoverer 2.1 was further analysed in Microsoft Office Excel, Graphpad Prism 9 and Perseus 1.6.10.43 using log2-transformed scaled total protein abundance data. A t-test was used to determine significantly differentially expressed proteins. Hierarchical clustering was performed in Perseus by Euclidean distance using k-means pre-processing. Gene ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) and enrichments were performed in Perseus using a Fisher test of significantly changed proteins versus the unchanged population after annotation with the proteins.

For Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) enrichment analysis, the ciDKO expression dataset was submitted to STRING with log2 fold change values associated with each protein. Functional enrichments were identified, and enrichment scores calculated based on aggregate fold change or Kolmogorov-Smirnov testing in STRING.

Long read RNA sequencing

RNA was isolated from ciDKO podocytes (3 days after Cre lentiviral transduction) and wild-type control cells (n=4 for each group) using an RNeasy mini kit (Qiagen). RNA was quantified using a Qubit Fluorimeter (Thermo Fisher Scientific) and quality assessed using the TapeStation system (Agilent). All samples had an RNA integrity number >8.

Barcoded PCR-cDNA library preparation was performed on 75 ng of RNA from each sample using the SQK-PCB109 kit (Oxford Nanopore Technologies). The resulting library was quantified using a Qubit Fluorimeter (Thermo Fisher Scientific) and molecular weight estimated using the TapeStation system (Agilent). 50 fmol of library was then loaded onto three promethION R9.4.1 flowcells (Oxford Nanopore Technologies), sequenced for 72 h and basecalled using guppy_basecaller 5.0.17 (Oxford Nanopore Technologies) in high accuracy mode.

Long read RNA analysis

IsoQuant analysis: the reads were aligned using IsoQuant (version 2.3.0), with option --data_type nanopore. The mouse gencode annotation (V29) (28) was used. The IsoQuant reference-based transcripts counts were used for further analysis. Differential expression analysis between the 2 conditions was performed using limma function with voom approach from limma Bioconductor package (29). The transcripts were analyzed without being grouped by genes, which will lead to a more stringent analysis than if done by gene (multiple correction testing is done by transcript). For visualization in the genome browser, the bam files were converted to bigWig files (using a scaleFactor calculated with edgeR (30)). FLAIR analysis (flair version 1.6.4): pychopper (version 2.7.2) (31) was first run on the total reads. Then the full length and rescued reads were aligned with “flair align” on the reference fasta file. The alignments were corrected using “flair correct”. The alignments were then collapsed using “flair collapse”, using the 8 replicates, into 1 complete reference. Then flair quantify, diffExp and diffSplice were run using the 2x4 replicates to get the rank lists of the 4 splicing event types (cassette, 3’, 5’, IR). A custom R script was used to find the nearest gene to the splicing events. In parallel, flair mark_intron_retention and predictProductivity (option --longestORF) were run on each corrected alignments to list the intron retention events by sample and predict the transcripts productivity by sample.

Quantitative PCR

cDNA was synthesised from RNA isolated from ciDKO and wild-type podocytes using the High-Capacity RNA-to-cDNA kit (ThermoFisher Scientific). Quantitative PCR was performed using SYBR Green JumpStart Taq ReadyMix (Merck) in a StepOnePlus PCR machine (Applied Biosystems). Primers with the following sequences were purchased from Life Technologies: Fn1 F 5’ CCCAGCTCACTGACCTAAGC 3’; Fn1 R 5’ GGAAGAGTTTAGCGGGGTCC 3’; Hcfc1r1 F 5’ GCCACCACTGGGGTAACTC 3’; Hcfc1r1 R 5’CTTCGGGAAAAGTCACAGGG3’; beta actin F 5’ GACAGGATGCAGAAGGAGATTACT 3’; beta actin R 5’ TGATCCACATCTGCTGGAAGGT 3’.

Statistical analysis

Statistical analysis was performed in Graphpad Prism 9 and data presented as the mean +/-SEM unless otherwise stated. Statistical significance was calculated with t-tests to compare two groups or one-way ANOVA to compare more than two groups and taken as p<0.05.

Data availability

The authors declare that all data supporting the findings of this study are available within the article and its supplementary information files or from the corresponding author upon reasonable request. The mass spectrometry proteomics data have been deposited in the PRIDE repository as part of the ProteomeXchange Consortium(32) under the dataset identifier PXD051018.

Results

podDKO mice develop a severe renal phenotype by 24 weeks

To study the effect of the combined loss of podocyte IR and IGF1R in vivo, transgenic mice were generated by crossing IRfl/fl/IGF1Rfl/fl mice with mice expressing Cre recombinase under the control of a podocin promotor (Figure S1A). There was no difference in body weight or blood glucose in podDKO mice compared with littermate controls (Figures 1A and B). However, knockout mice developed albuminuria with significantly increased urinary albumin:creatinine ratio (uACR) (p<0.01) at 24 weeks of age (Figure 1C). The severity of the phenotype was variable and while some mice exhibited only mild albuminuria, more severe renal disease was observed in others resulting in death in ∼20% of the mice. uACR was not increased in Cre negative and podCre expressing mice at 6 months excluding Cre toxicity as the driver of albuminuria in podDKO mice (Figure S1C). Glomerulosclerosis was significantly increased (p<0.05) in knockout mice with areas of sclerosis and tubular protein casts visible on periodic acid-Schiff (PAS) stained sections (Figure 1D), together with evidence of increased fibrosis as determined by Masson’s trichrome staining (Figure 1E). Transmission electron microscopy (TEM) analysis revealed disruption of the filtration barrier ultrastructure with significantly increased (p<0.01) foot process width (Figure 1F). Podocyte number, evaluated by immunofluorescence (IF) using a podocyte specific WT1 antibody, was significantly (p<0.0001) reduced in podDKO mice (Figure 1G).

Simultaneous knockout of podocyte IR and IGF1R in vitro is highly detrimental

We surmised that the variable phenotype observed in our transgenic mouse model was due to differences in the efficiency of Cre mediated excision and level of receptor knockdown. With the aim of developing a model with near total loss of the IR and IGF1R to use in in vitro studies, a conditionally immortalised podocyte cell line was generated from IRfl/fl/IGF1Rfl/fl mice and these cells transduced with Cre expressing lentivirus (ciDKO) (Figure S2). This allowed us to initially culture the cells containing both receptors before their simultaneous deletion using Cre recombinase. Western blot analysis showed a decrease of the IR (p<0.0001) and IGF1R (p<0.0001) in ciDKO podocytes of >80% (Figure 2A). Phosphorylation of AKT and p44/42 MAPK in response to either acute insulin or IGF1 stimulation was reduced (p<0.05) indicating suppression of insulin/IGF1 signalling in ciDKO podocytes (Figures 2B and C). Compound loss of both the IR and IGF1R was detrimental, resulting in >50% podocyte loss (p<0.0001) 7 days after Cre-mediated gene excision (Figure 2D).

Simultaneous knockout of podocyte IR and IGF1R in vitro is highly detrimental

A. Western blot shows >80% reduction of IR and IGF1R protein in ciDKO cells. t-test, ****p<0.00001, n=18.

B. Phosphorylation of AKT and p44/42MAPK in response to acute insulin stimulation at 10 nM and 100 nM for 10 minutes was significantly reduced in ciDKO podocytes. One way ANOVA, **p<0.01, *p<0.05, n=3.

C. Phosphorylation of AKT and p44/42MAPK in response to acute IGF1 stimulation at 10 and 100 ng/ml for 10 minutes was reduced in ciDKO podocytes. One way ANOVA, *p<0.05, n=3.

D. Fewer than 50% of ciDKO cells survive 7 days after gene excision. t-test, ****p<0.0001, n=3-4 independent experiments.

Proteomic analysis of ciDKO podocytes reveals downregulation of spliceosome proteins

To elucidate the mechanistic processes regulated by insulin/IGF1 signalling in podocytes, we performed unbiased tandem mass tagged (TMT) LC MS/MS proteomics to compare total protein expression in ciDKO compared with wild-type cells (Figure 3A). Hierarchical clustering and principal component analysis (PCA) was performed using Perseus software to determine that samples were clustered according to the experimental group (Figures S3A and S3B). 4842 proteins were significantly differentially expressed in ciDKO cells compared to wild-type controls with 2127 proteins upregulated and 2715 proteins downregulated (Figure 3A and Figure S3C). Perseus software was used to perform hierarchical clustering and identify Gene Ontology (GO) and Kyoto Encyclopaedia of Genes and Genomes (KEGG) terms enriched in ciDKO podocytes (Figure 3B). This revealed a downregulation of proteins involved in DNA replication and cell cycling along with a striking downregulation of proteins involved in spliceosome function and RNA processing (Figure 3B). These findings were confirmed using Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) to identify KEGG terms overrepresented in the ciDKO proteome (Figure 3C). STRING identified the spliceosome as the most significantly enriched KEGG term in dual receptor knockout cells with a -log10 false discovery rate ∼2 fold greater than any other (Figure 3D).

Proteomic analysis of ciDKO podocytes reveals downregulation of spliceosomal proteins

A. Schematic outlining workflow for proteomic analysis.

B. Heat map showing hierarchical clustering of ciDKO vs wild-type podocyte proteomes (decreased protein expression in green, increased expression in red) and the GO and KEGG terms enriched in four major clusters.

C. and D. KEGG enriched terms in STRING. Downregulated pathways (green) are associated with higher enrichment scores in comparison to upregulated pathways (red). Enrichment scores are computed by STRING using the Kolmogorov-Smirnoff test. KEGG term “Spliceosome” (highlighted border) is associated with a high enrichment score and the highest false discovery rate across all terms.

E. Western blots show significantly reduced levels of EIF4A, SF3B4 and PTBP2 in ciDKO podocytes compared with wild-type cells. Unpaired t-test, ***p<0.001, **p<0.01, *p<0.05, n=3 independent experiments.

To validate results from the proteomic data, western blot analysis of wild-type and ciDKO podocytes was performed using antibodies to selected spliceosome proteins. This confirmed that levels of eukaryotic initiation factor-4A (EIF4A) (p<0.001), splicing factor 3B subunit 4 (SF3B4) (p<0.01) and polypyrimidine tract binding protein 2 (PTBP2) (p<0.05) were significantly reduced in knockout cells (Figure 3E).

Exposure of cultured podocytes, but not glomerular endothelial cells, to the spliceosome inhibitor pladienolide B results in dose dependent cell death

Proteomic analysis indicated a significant reduction in spliceosome proteins associated with all stages of the splicing process in ciDKO podocytes (Figure 4A). We therefore hypothesised that impaired spliceosome assembly and function may be at least partially responsible for the deleterious phenotype observed with podocyte IR/IGF1R knockout. To determine the importance of spliceosome activity to mature podocyte homeostasis and to ascertain whether different cell types are equally sensitive to spliceosome inhibition, the cytotoxic effect of pladienolide B (a spliceosome inhibitor targeting SF3B) (33) was compared in HeLa cells, fully differentiated, wild-type murine podocytes and glomerular endothelial cells. Consistent with previous observations (33), exposure of HeLa cells to pladienolide B at 0-100 nM for 48 hours was cytotoxic at ≥1 nM (Figure 4B). Pladienolide B-induced spliceosome inhibition also resulted in a significant (p<0.05) dose dependent loss of podocytes at a concentration of ≥25 nM (Figure 4D and E). In contrast, no cytotoxicity was observed in glomerular endothelial cells (Figure 4C).

Exposure of cultured podocytes, but not glomerular endothelial cells, to the spliceosome inhibitor pladienolide B results in dose dependent cell death

A. Annotated KEGG pathway showing a majority of spliceosome proteins are significantly downregulated (green) in ciDKO podocytes.

B-D. Pladienolide B exposure for 48 hours in HeLa cells (B), glomerular endothelial cells (GenC) (C) and podocytes (D). One way ANOVA, ****p<0.0001, **p<0.01, *p<0.05, n=3 independent experiments.

E. Bright field images of wild-type podocytes exposed to the indicated concentrations of pladienolide

B. Scale bar=100 μm

Loss of podocyte IR and IGF1R is associated with increases in intron retention and unproductive transcript expression

To obtain an overview of global differences in various alternative splicing events in the ciDKO transcriptome, long-read RNA sequencing was used. Long-read sequencing was preferred as short-read RNA seq tools have been found to detect such events with poor accuracy (34). RNA was isolated from ciDKO and wild-type control cells, sequenced using long-read RNA Oxford Nanopore PromethION flow cells (Figure 5A) and the resulting data analysed using the Full-Length Alternative Isoform Analysis of RNA (FLAIR) tool. An overview of the main types of alternative splicing events is depicted in Figure 5B and rank lists of genes in ciDKO podocytes with significantly increased (adjusted p value <0.05) intron retention splicing; alternative 5’ splicing; alternative 3’ splicing and cassette exon splicing are shown in Supp Tables 1-4. KEGG pathway enrichment analysis using STRING, shows an overlap in overrepresented pathways (including Spliceosome, RNA transport and Ribosome) from rank lists of genes with intron retention, cassette exon splicing and alternative 5’ splicing (Supp Figure 4D-F).

Loss of podocyte IR and IGF1R is associated with increases in intron retention and unproductive transcript expression

A. Schematic outlining long read RNA sequencing workflow

B. An overview of the alternative splicing events quantified by FLAIR. Boxes represent exons (blue=constitutive exons; yellow=alternative exons), lines represent introns.

C. Boxplot showing the fraction of transcripts with intron retention events in ciDKO and wild-type podocytes. t-test, **p<0.01, n=4 for each experimental condition.

D. Box plot of the fraction of productivity events in the ciDKO and wild-type transcriptomes shows a higher proportion of unproductive transcripts in ciDKO podocytes. NGO=no start codon; NST=start codon but no stop codon; PRO=productive transcripts; PTC=premature termination codon ie. unproductive transcripts. t-test, ***p<0.001, **p<0.01, *p< 0.05, n=4 for each experimental condition.

E. Proportional stacked bar graph of productivity events in the ciDKO and wild-type transcriptomes. NGO=no start codon; NST=start codon but no stop codon; PRO=productive transcripts; PTC=premature termination codon ie. unproductive transcripts.

To compare the intron retention in the ciDKO versus controls, the number of intron retention events by sample were counted using FLAIR mark_intron_retention, and then divided by the number of transcripts per sample. The ciDKO samples have a significantly (p-value = 0.004) higher proportion of transcripts with intron retention: ∼18% of transcripts in ciDKO podocytes compared with 14% in wild-type controls (Figure 5C). The overlap of the intron retention events between experimental repeats is ∼30% in both wild-type and ciDKO podocytes. The transcript productivity was also assessed using FLAIR predictProductivity with transcripts classified in 4 categories: PRO (productive); PTC (premature termination codon, i.e. unproductive); NGO (no start codon) and NST (has start codon but no stop codon appended to the end of the isoform name). The fraction of transcripts harbouring a premature termination codon is significantly higher (p<0.001) in ciDKO podocytes (Figure 5D and E).

Compound knockdown of the IR and IGF1R alters the podocyte transcriptome

To determine the effect of spliceosome depletion on the transcriptomic profile of ciDKO podocytes, IsoQuant (35) was used to quantify genes and transcripts from existing GENCODE mouse annotation, then limma (29) used to determine differential gene and transcript expression. The number of reads per sample is shown in Figure S4A. Variance-stabilized transformed RNA Seq data (Figure S4 B) were used in principal component analysis and confirmed that samples were clustered according to the experimental group (Figure S4C). The number of minimally expressed transcripts in the experiment was 31580 and using an adjusted p-value threshold of 0.01, 831 transcripts were significantly upregulated and 570 significantly downregulated in ciDKO podocytes. The top 10 most significantly differentially expressed genes (analysed by transcript) are shown in Table 1. These included down-regulation of host cell factor C1 regulator 1 (Hcfc1r1), a nuclear export factor which regulates the activity of the transcription factor host cell factor 1 (Hcf1) and which analysis using FLAIR identified as subject to increased alternative 3’ splicing in ciDKO podocytes (Supp Table 4). Also, fibronectin 1 (Fn1) encoding an extracellular matrix protein which exists as different isoforms derived by alternative splicing at three sites: Extra Domains A and B exons (EDA and EDB) and the type III connecting segment (36). Fn1 splice variants were differentially expressed in ciDKO podocytes with increased expression (log fold change of 1.11[=2.16 fold]) of an Fn1 transcript containing the EDA and EDB exons (Figure 6A-C). qPCR validation confirmed the decrease in Hcfc1r1 expression (Figure 6D) while primers designed to specifically amplify transcripts containing the fibrosis-associated Fn1 EDB exon also verified that expression of such transcripts is significantly increased (1.8 fold, p<0.001) in ciDKO podocytes when compared with wild-type control cells (Figure 6E).

Compound knockdown of the IR and IGF1R alters the podocyte transcriptome

A. UCSC genome browser tracks of transcripts annotated to Fn1. Red lines indicate the number of reads in each sample containing the EDA and EDB exons.

B. Boxplot showing differential expression of an Fn1 transcript containing EDA/EDB exons in ciDKO podocytes.

C. Exon/intron structure of Fn1 transcript differentially expressed in IGF1R DKO vs wild-type podocytes. Expression of an EDA and EDB exon-containing transcript (not normally expressed in mature podocytes and associated with fibrosis) is increased in IGF1R DKO podocytes.

D. qPCR using primers specific to the Fn1 EDB exon shows an increase in the expression of fibrosis associated EDB exon-containing transcripts in ciDKO podocytes. t-test, **p<0.001, n=4 independent experiments.

E. qPCR shows decreased expression of Hcfc1r1 mRNA in ciDKO podocytes. t-test, ****p<0.00001, n=4 independent experiments.

Discussion

We have previously shown that signalling to the podocyte via the IR is critical for kidney function (17) while the IGF1R has a role in maintaining podocyte mitochondrial function (21). The current study used in vivo and in vitro models lacking both receptors simultaneously to further elucidate important common signalling pathways regulated by the insulin/IGF1 axis in the podocyte.

We initially used a transgenic mouse model (podDKO) to determine the physiological consequences of dual receptor loss in vivo. These mice developed an obvious renal phenotype: significant albuminuria, glomerulosclerosis, and ultrastructural filtration barrier damage was apparent by 6 months and was severe enough to cause death in a number of mice. However, body weight and blood glucose in podDKO mice were comparable to controls, in contrast with several other mouse models where alongside tissue-specific manifestations, double receptor knockout also gives rise to systemic effects. For example, IR/IGF1R knockout in kidney proximal tubule cells (10) and muscle (11) both result in decreased body size while other models have impairments in systemic metabolism: mice with adipose tissue-specific IR/IGF1R knockout develop diabetes (9) while dual receptor knockout in hippocampus and central amygdala results in a decrease in glutamate receptors and reduced glucose tolerance (12).

We performed unbiased proteomic analysis to identify the mechanistic processes responsible for the detrimental phenotype observed in our in vivo and in vitro models. Given the well-established mitogenic effect of IGF1 signalling (37), this showed a predictable downregulation of cell cycle and DNA replication pathways. However, we also observed a decrease in DNA damage response pathways and histone methylation. As post-mitotic cells with a limited capacity for replenishment, maintaining genomic integrity is crucial for podocyte health (3840). Furthermore, altered histone methylation, another feature of the podocyte IR/IGF1R knockdown proteome, has been shown to affect the function of the DNA repair system in podocytes (41). Recent studies suggest that podocyte DNA damage and changes in glomerular DNA methylation may be useful markers of chronic kidney disease and correlate with estimated glomerular filtration rate (eGFR) decline in patients with diabetic nephropathy (42). It therefore seems plausible that deficiencies in DNA damage repair mechanisms contribute to the detrimental phenotype and ultimate loss of podocytes resulting from IR/IGF1R knockdown.

Evidence from previous studies suggests that DNA damage is intrinsically linked with spliceosome function: DNA damage affects the expression, activity and localisation of splicing factors (43) while spliceosome dysfunction leads to DNA damage by promoting the aberrant formation of R loops during transcription (44, 45). Thus, accurate spliceosome function is critical for maintaining genomic integrity (44). In addition to the decreased expression of DNA damage response proteins revealed by our proteomic data, we also observed a striking reduction in the levels of proteins required for spliceosome assembly and regulation in ciDKO podocytes. We have previously used lentiviral mediated expression of Cre recombinase to induce knockout of other podocyte genes, notably glycogen synthase kinase 3 α and β in combination (27). While this resulted in a highly detrimental cellular phenotype, proteomic analysis of these cells showed no differences in the expression of spliceosome related proteins, so we are confident that the results from the current study are not due to generalised cell death/dysfunction.

The spliceosome is a large ribonucleoprotein complex whose critical function in all eukaryotic cells is to catalyse the removal of introns from precursor messenger RNA molecules and join the flanking exons to generate mature mRNA transcripts that direct protein synthesis during translation (46). Spliceosome function is also a driver of significant proteomic diversity via alternative splicing, whereby multiple mRNA transcripts containing different exon combinations are produced from a single gene (46). Mutations in core spliceosome components, spliceosome assembly factors or regulatory proteins can result in defects in pre-RNA processing, increased nonsense-mediated decay of affected transcripts, abnormal or truncated protein products and changes to alternative splicing. Such erroneous splicing can substantially alter the transcriptome in specific cell types. This has been shown in immune cell dysfunction (47), cancer (47) and neurodegeneration (48, 49). Somatic mutations in genes encoding spliceosome components can result in diseases, known as spliceosomopathies, (including retinitis pigmentosa, myelodyplastic syndromes and craniofacial disorders (50)), with tissue-specific manifestations, suggesting that although spliceosome factors are ubiquitously expressed, they likely have cell type specific functions. Interestingly, one of the splice factors frequently involved in these spliceosomopathies - SF3B4 - is down regulated in the double knock-down podocytes (Figure 3E). We used the spliceosome inhibitor pladienolide B (51), which targets the SF3B complex, to investigate the effect of reduced spliceosome activity in wild-type podocytes. Developed as an anti-neoplastic agent, pladienolide B is effective at reducing cell viability, particularly in cancer cell lines (5154), presumably due to an increased demand for translation and protein production in such actively dividing cells. In contrast, a comparatively high dose of pladienolide B is not sufficient to induce cytotoxicity in HEK 293T cells, suggesting that not all cell types are equally sensitive to spliceosome inhibition (55). While we were able to reproduce the low dose cytotoxic effect of this spliceosome inhibitor in HeLa cells reported previously (51), the response of the glomerular cell types we assayed differed. No cytotoxic effect was apparent in glomerular endothelial cells while in contrast, we found podocytes to be sensitive to pladienolide B, resulting in significant cell death and suggestive of the importance of adequate spliceosome function in this cell type. The splicing factor CD2 antigen cytoplasmic tail binding protein 2 (CD2BP2) is known to be essential for kidney development and podocyte integrity (56) while alternative splicing of genes including Wilm’s tumour gene 1 (WT1) (57) and vascular endothelial growth factor A (VEGF-A) (58) is important in podocyte homeostasis and disruption of this tightly regulated process associated with chronic kidney disease (59). We therefore performed long-read RNA sequencing to explore the effect of podocyte IR/IGF1R loss on spliceosome function and the transcriptome of ciDKO cells.

Long read RNA Seq analysis shows that podocyte IR/IGF1R depletion has a marked impact on alternative splicing events, notably an increase in the fraction of transcripts with intron retention/premature termination codons. Interestingly, enrichment analysis shows that many transcripts associated with spliceosome function and RNA transport have increased levels of intron retention, cassette exon splicing and alternative 5’ and 3’ splicing. Although it has recently been recognised that intron retention can serve as a mechanism for regulating gene expression (60) including the regulation of RNA processing genes (61), our results could also be explained by the apparent loss of the spliceosome in ciDKO podocytes. Indeed, intron retention secondary to mutations in spliceosome components is a hallmark of a number of cancers (62) and neurodegenerative conditions (49). Moreover, it is likely that our RNA seq analysis underestimates the incidence of transcripts with premature termination codons as many will be targeted for nonsense-mediated decay (63) and would thus be unmeasurable in our samples.

Significant differential transcript expression was apparent in ciDKO podocytes with transcripts encoding the extracellular matrix protein fibronectin among the top hits. We detected increased expression of a Fn1 splice variant which included the EDA and EDB exons. Although EDA/EDB-fibronectin is important during development, its expression is limited in adult tissue including kidney (64). However, EDA-fibronectin is upregulated in cultured podocytes exposed to TGFβ treatment (65) and a pathological increase in this isoform results in insulin resistance (66), together with fibrosis in multiple tissues including lung (67) and kidney (68). It is conceivable that the changes in podocyte splicing resulting from IR/IGF1R knockdown promote a fibrotic phenotype by altering the ratio of fibronectin variants expressed, consistent with the increased fibrosis observed in our podDKO mice.

In summary, we have shown that the IR/IGF1R axis is critical for podocyte health. Mechanistically it does this through controlling gene transcription by maintaining spliceosome function. This suggests hormonal signals are important for the genetic control of these terminally differentiated cells.

Acknowledgements

We thank the Proteomic Facility and Histology Services Unit at the University of Bristol for technical support. This study was funded by the Medical Research Council (Senior Clinical Fellowship) to RJMC -(MR/K010492/1), Kidney Research UK, and funding from the Innovative Medicines Initiative 2 Joint Undertaking (JU) under grant agreement No 115974. The JU receives support from the European Union’s Horizon 2020 research and innovation programme and EFPIA and JDRF. Any dissemination of results reflects only the author’s view; the JU is not responsible for any use that may be made of the information it contains.

Additional information

Author contributions

The study was conceived by R.J.M.C., G.I.W. and J.A.H. Latterly academic input was supplied by M.H. and P.T.B. Experiments were performed by J.A.H., L.D., J.T.C. and F.B. Conditionally immortalised cell lines were made by L.N. PodCre mice were provided by P.T.B. IGF1R fl/fl mice were provided by M.H. Long read RNA sequencing was performed and processed by A.J. and S.B., F.B. bioinformatically analysed and interpreted the data. The manuscript was written by J.A.H and R.J.M.C. and all authors reviewed, commented and approved the paper.

Additional files

Supplementary figures

Supp Table 1

Supp Table 2

Supp table 3

Supp Table 4